Zen gets even with Broadwell-E in Blender

[Shintai]

Zen doesn't need to be as fast as top end Xeons to be successful, as long as it gets to Ivy Bridge levels margin improvements will pull AMD into the black. The foundation of Zen if its good enough to do that, then they will stabilize future versions as well. Of course we still don't know what Intel will do, but that is too far ahead to really get any clear picture of that.

There is a few cave rats, because the datacenter business is ruthless in this point. If Zen cant compete with Xeons in terms of perf/watt. Then sales will be ~0 in the server segment.

Zen if it reaches Ivy bridge levels or Haswell, AMD will have a 3 year gap where they can make profits, but that is only if Intel sits idle!, AMD has already shown their cards when it comes to future projections of Zen so Intel can plan ahead for their next CPU's, Cannon Lake and make sure AMD doesn't get anywhere. So that cuts that three years down to a year and half. This is all with only margins increases and slight marketshare uptake in desktop and notebook.

At 3Ghz? Certainly not. At same clocks? Highly doubtful. Zen isn't going to make AMD profits, unless its actually beats or equals competition. If AMD is going to sell top parts for cheap for them to sell at all. Then its not making any profit to write home about. Also they need more revenue. And the Zen APU is a Q4 2017 if not later product. 99% of Intels consumer sales is CPUs with IGP.

If the Zen quad core is a 2.8/3.2Ghz part as information so far points to. Then its already a sub 150$ product. And even if they reach Haswell IPC, its to fight against something like i5 4430 that was sold 3 years earlier.

Zen is another dream from AMD to reenter the server segment. Something way out of reach. Desktop and mobile is all secondary thoughts.

 

[NKD]

Why dont you tell me why Zen will be good instead of playing the hate card for the nth time and act like the victim? What miracles will AMD do on their shoehorn R&D budget to make it better? They even got less R&D for it than Bulldozer had. Why did AMD issue stocks now, if Zen is going to be this great competitive product that will just reap in revenue in half a year? Why not show Zen left and right if it is so good when there is nothing to lose but everything to win from it? How is 1 AMD engineer going to compete with ~10 Intel engineers that uses a much better node and got the IDM benefit?

There is a reality to stick to, even when you root for the underdog with all your desires. Just as well as VIA isn't going to make a competitive higher end x86 product anymore either.

why dont you show me the processor in your hand and running cinebench or something and give me all the benchmarks. Until you do you are no more right than I am. Cuz unlike Razor and other people who actually take more reasonable approach to this despite their criticism of zen. If Zen can be competitive in IPC with in 20% what do they have left clock speeds given that its 3.0 or 3.2? You really think they can't ramp those up with new revisions every year or so with zen+?

You take my way or the high way approach. AMD doesn't have to beat intel, because they just have to become competitive first and price the product right. Thats all they need, if they fail at that then yea you may be right. But you don't know that anymore than I do or anyone else. No I don't have benchmarks of zen to tell you why it will be good. Because I rather wait until the product is out and then judge it.

The point you fail to realize is I am not rooting for anyone. I am just taking the wait and see approach. I am not going to waste my time on a product that has no real benchmark out. Where did VIA come from? lol. You keep saying the same thing over and over when I say I wanna wait and see where the product lands. Root for underdog? I don't throw my money on products that I don't value and I don't have much that I value from AMD at this point other than rx 470, could zen change my opinion? May be, but I can't say that until I see the product launched and reviewed. Afterall I am running a 6700k. I am usually the person who likes to give people/company benefit of the doubt, and I doing the same with zen. If its a failure well so be it! It won't hurt my feelings, but if it shines good for the market. Because I don't think there is one post I have seen from you that is remotely positive about amd. You said polaris was a failure, but I think rx470 is probably the best card at 200 that I have ever seen. Rx 480? not much. Rx 470 makes rx 480 look like a bad buy to me. So polaris is not a failure, whats failure is not having a competing product on high end.

Like I said I will wait and see. I don't have to show you anything because I don't even know if it will be any good but I am not going to say its a failed product before seeing real benches.

 

[razor1]

There is a few cave rats, because the datacenter business is ruthless in this point. If Zen cant compete with Xeons in terms of perf/watt. Then sales will be ~0 in the server segment.

Agreed, they won't get anything in the server market if they can't compete with Sky Lake Xeon's.

At 3Ghz? Certainly not. At same clocks? Highly doubtful. Zen isn't going to make AMD profits, unless its actually beats or equals competition. If AMD is going to sell top parts for cheap for them to sell at all. Then its not making any profit to write home about. Also they need more revenue. And the Zen APU is a Q4 2017 if not later product. 99% of Intels consumer sales is CPUs with IGP.

If the Zen quad core is a 2.8/3.2Ghz part as information so far points to. Then its already a sub 150$ product. And even if they reach Haswell IPC, its to fight against something like i5 4430 that was sold 3 years earlier.

Zen is another dream from AMD to reenter the server segment. Something way out of reach. Desktop and mobile is all secondary thoughts.

even at 3 ghz they can, as long as power consumption stays within the target ranges.

The reason being, right now they still have 20% or so marketshare in desktops and notebooks, that marketshare with 10-15% increased margins is enough to make a quite a bit of money, The margins will increase automatically

a) die size will be considerably smaller than their previous chips
b) performance will be higher than their current chips.

b will help them put a lower core part against Intel's current line up. Instead of a 8 core vs. 4 core i3, i5, they will have a 4 core vs 4 core Intel possibly at lower clocks, because b as well, ASP's go up.

Both those factors together. You can see that they will start making profits even at Ivy bridge level IPC.

Anything lower than Ivy Bridge though, they will be stuck in the same rut, they are in now. Hopefully this isn't the case, but we have seen AMD pull the rug over our eyes more than once. The turn around to this is, so far all the papers and patents and everything they have shared this far, it seems like Zen will be a good product.

I don't really care about their presentations and hype about who made what.

From AMD's point, this is what they need, coming back after being hit so hard for so long, they are going to have to start off with small gains. This isn't like the P4 days, Intel has not made any missteps on their architectures so we can't expect AMD is just to trounce on Intel's offerings like with Athlon. So this reprieve for at least 1.5 years to 3 years, they need to make the best use of their time and money and go in the right direction, because if they don't, Intel will tighten the noose, AMD's problems aren't over yet, this is just the first step to ease the pain they have been in, there needs to be more steps to fully get out of their hole, which is their debt. First payable will come up in 2019, so that cash, has to be in hand and stay there, which they have now.

Any incoming profits from Zen will keep them going with what they are doing now, unlike in the past where they were burning capital. So even expanding into new markets isn't really going to be on their top priority till that debt is cleared or they really gain momentum on both the GPU and CPU sides.

 

[Shintai]

Agreed, they won't get anything in the server market if they can't compete with Sky Lake Xeon's.


even at 3 ghz they can, as long as power consumption stays within the target ranges.

The reason being, right now they still have 20% or so marketshare in desktops and notebooks, that marketshare with 10-15% increased margins is enough to make a quite a bit of money, The margins will increase automatically

a) die size will be considerably smaller than their previous chips
b) performance will be higher than their current chips.

b will help them put a lower core part against Intel's current line up. Instead of a 8 core vs. 4 core i3, i5, they will have a 4 core vs 4 core Intel possibly at lower clocks, because b as well, ASP's go up.

Both those factors together. You can see that they will start making profits even at Ivy bridge level IPC.

Anything lower than Ivy Bridge though, they will be stuck in the same rut, they are in now. Hopefully this isn't the case, but we have seen AMD pull the rug over our eyes more than once. The turn around to this is, so far all the papers and patents and everything they have shared this far, it seems like Zen will be a good product.

I don't really care about their presentations and hype about who made what.

From AMD's point, this is what they need, coming back after being hit so hard for so long, they are going to have to start off with small gains. This isn't like the P4 days, Intel has not made any missteps on their architectures so we can't expect AMD is just to trounce on Intel's offerings like with Athlon. So this reprieve for at least 1.5 years to 3 years, they need to make the best use of their time and money and go in the right direction, because if they don't, Intel will tighten the noose, AMD's problems aren't over yet, this is just the first step to ease the pain they have been in, there needs to be more steps to fully get out of their hole, which is their debt. First payable will come up in 2019, so that cash, has to be in hand and stay there, which they have now.

Any incoming profits from Zen will keep them going with what they are doing now, unlike in the past where they were burning capital. So even expanding into new markets isn't really going to be on their top priority till that debt is cleared or they really gain momentum on both the GPU and CPU sides.

Quad 3Ghz(2.8/3.2) with less IPC at 65W without IGP. See the issue? Its not going to sell for much. You are already down to performance 3-5 years ago and lacking the IGP.

A: Die size may be smaller. But is the cost less? Remember transistor cost didn't go down. And IC design cost went up. So I am willing to bet that each Zen CPU/APU will cost them more than the equal construction CPU/APU.
B: Even with the improvements to the construction core line, they didn't bother with a CPU only version because the market is simply too small. And remember, the Zen APU is really far away(10nm Cannon Lake time). If the server adventure fails, then they lack this part. And a IGP less CPU isn't going to sell much. ~0 in laptops.

So you cant assume automated profit or improved margins.

Zen is designed for servers first of all, just like Bulldozer and Phenom. And that's one of the troubles. The nemesis if you like of the company. Same reason the RTG group got sacrificed.

AMDs big problem is its R&D. You just dont get something you didn't pay for. And that issue becomes even bigger when you try and go for the unreachable (Servers).

 

[Pieter3dnow]

So you cant assume automated profit or improved margins.

Zen is designed for servers first of all, just like Bulldozer and Phenom. And that's one of the troubles. The nemesis if you like of the company. Same reason the RTG group got sacrificed.

AMDs big problem is its R&D. You just dont get something you didn't pay for. And that issue becomes even bigger when you try and go for the unreachable (Servers).

But you can assume all the things you say. Don't think so.

You talk about R&D but it is not R&D as the only demo of Zen out there shows it is rather good. What we don't know are the clock speeds of the retail product will be and that is not down to R&D is it.

Nowhere is it mentioned that Zen is an exclusive design for servers , K12 is ...

If server market is unreachable then AMD wasted another 3 years and more money than they can afford, that would not make any sense to run a business like that.

 

[KazeoHin]

My wound from Bulldozer still hurts. I won't be led to have faith in any AMD product before it launches.

Is there a chance that Zen is awesome? Of course! But bench it, and I'll believe it. No bench, no faith.

 

[Shintai]

But you can assume all the things you say. Don't think so.

You talk about R&D but it is not R&D as the only demo of Zen out there shows it is rather good. What we don't know are the clock speeds of the retail product will be and that is not down to R&D is it.

Nowhere is it mentioned that Zen is an exclusive design for servers , K12 is ...

If server market is unreachable then AMD wasted another 3 years and more money than they can afford, that would not make any sense to run a business like that.

What if that demo is as trusty as the Polaris demo? Or the Bulldozer? Or the Phenom? AOTS and GB4 leak is far from those results to put it mildly.

K12 is gone.

Phenom and Bulldozer for the server market didn't make sense either. But yet it happened. And AMD haven't lost more money than they can afford yet. Long from it.

 

[razor1]

Quad 3Ghz(2.8/3.2) with less IPC at 65W without IGP. See the issue? Its not going to sell for much. You are already down to performance 3-5 years ago and lacking the IGP.

A: Die size may be smaller. But is the cost less? Remember transistor cost didn't go down. And IC design cost went up. So I am willing to bet that each Zen CPU/APU will cost them more than the equal construction CPU/APU.
B: Even with the improvements to the construction core line, they didn't bother with a CPU only version because the market is simply too small. And remember, the Zen APU is really far away(10nm Cannon Lake time). If the server adventure fails, then they lack this part. And a IGP less CPU isn't going to sell much. ~0 in laptops.

So you cant assume automated profit or improved margins.

Zen is designed for servers first of all, just like Bulldozer and Phenom. And that's one of the troubles. The nemesis if you like of the company. Same reason the RTG group got sacrificed.

AMDs big problem is its R&D. You just dont get something you didn't pay for. And that issue becomes even bigger when you try and go for the unreachable (Servers).

Their current crop of CPU's are on 32 nm, going to 14nm, is going to save them money lol, if you follow the trend of nodes, the cost flattens out as the process matures. 14nm might be more expensive per transistor, but for mm2 over 32nm (comparing same performance chips) its cheaper.

R&D isn't the only defining factor of creating a good product. If you look back into the early days of AMD, they didn't have the R&D budget Intel had, still they were able to do well. Granted they reverse engineered Intel's chips, but still, a smaller budget got them to equal stance with Intel.

Their knowledge of what went wrong with Bulldozer and Phenom, just by fixing those problems they can get a core that can compete somewhat. Its not about reinventing the wheel, they aren't that far behind.......

Also people buying these chips for gaming (low end gaming which I don't think that is the case), won't need an IGP (if this is the case) and right now, these are the only people that are buying AMD CPU's, no businesses buy AMD, at least smart ones, because they know just the power savings alone is enough to go to Intel.

 

[Pieter3dnow]

What if that demo is as trusty as the Polaris demo? Or the Bulldozer? Or the Phenom? AOTS and GB4 leak is far from those results to put it mildly.

You are in the blender demo thread guess what demo there was of Zen
The performance in the demo is tied to that AMD can compile the source, if you know this then you can expect that programs not compiled for Zen will do worse, so this is the best case scenario (nothing wrong with that).

K12 is gone.

Where did you read this ?
AMD K12 - Wikipedia, the free encyclopedia

Phenom and Bulldozer for the server market didn't make sense either. But yet it happened. And AMD haven't lost more money than they can afford yet. Long from it.

CRAYze people ?
Cray Announces AMD Bulldozer CPU and NVIDIA Tesla GPU Supercomputer Capable of 50 Petaflops | PC Perspective

 

[Shintai]

Their current crop of CPU's are on 32 nm, going to 14nm, is going to save them money lol, if you follow the trend of nodes, the cost flattens out as the process matures. 14nm might be more expensive per transistor, but for mm2 over 32nm its cheaper.

R&D isn't the only defining factor of creating a good product. If you look back into the early days of AMD, they didn't have the R&D budget Intel had, still they were able to do well. Granted they reverse engineered Intel's chips, but still, a smaller budget got them to equal stance with Intel.

Their knowledge of what went wrong with Bulldozer and Phenom, just by fixing those problems they can get a core that can compete somewhat. Its not about reinventing the wheel, they aren't that far behind.......

Also people buying these chips for gaming, won't need an IGP, and right now, these are the only people that are buying AMD CPU's, no businesses buy AMD, at least smart ones, because they know just the power savings alone is enough to go to Intel.

Cost of 14/16nm isn't going down until EUV comes. Dont count mm2, count transistors.

Cloning chips isn't so R&D intensive. And even at K8, Intel had P-M. Intel made all the mistakes in the server and desktop area. But P-M for example was unchallenged.

I think you vastly underestimate the effort needed for a core design.

I dont think many people buy AMD for gaming, and they sure need the IGP as well. Only something like 10-15% of all PCs comes with a dGPU.

 

[Shintai]

you are in the blender demo thread guess what demo there was of Zen


Where did you read this ?
AMD K12 - Wikipedia, the free encyclopedia



CRAYze people ?
Cray Announces AMD Bulldozer CPU and NVIDIA Tesla GPU Supercomputer Capable of 50 Petaflops | PC Perspective

It wasn't even a standard Blender run.

Wiki is quite useless. And the last info about K12 was almost 1½ year ago. Never mentioned since. Well, besides this:
AMD turns back to x86 for server reboot as it downgrades ARM

Cray had a contract with AMD, they ditched AMD as quickly as they could in favour of Xeons afterwards.
Cray jumped from AMD for the flexibility and performance of Intel chips | TheINQUIRER

 

[razor1]

CRAYze people ?
Cray Announces AMD Bulldozer CPU and NVIDIA Tesla GPU Supercomputer Capable of 50 Petaflops | PC Perspective

Cray doesn't sell those anymore lol.

They had them for like a few months and after that were replaced with Xeons.

Advanced Cluster Computing: CS400 and CS-Storm | Cray

Even AMD's name has been removed off their partners list.

 

[razor1]

Cost of 14/16nm isn't going down until EUV comes. Dont count mm2, count transistors.

Cloning chips isn't so R&D intensive. And even at K8, Intel had P-M. Intel made all the mistakes in the server and desktop area. But P-M for example was unchallenged.

I think you vastly underestimate the effort needed for a core design.

I dont think many people buy AMD for gaming, and they sure need the IGP as well. Only something like 10-15% of all PCs comes with a dGPU.

I agree design costs increase, but that doesn't affect margins, design costs, are not fixed costs, they are variable costs and should never be mixed in with margins.

Having stated that, the total R&D budget of Zen is not just one year, its probably 5 years ago when they started?

So they have enough money to make a decent chip. We aren't talking about beating Intel here, AMD knows why Intel's chips are beating them, they know what changes they have to make, its not something they need to "invent" they just need to figure out how to do it. Over a 5 year span and decent lead they should be able to do that. If they don't, well, AMD is dead. That is it.

 

[Pieter3dnow]

It wasn't even a standard Blender run.

Wiki is quite useless. And the last info about K12 was almost 1½ year ago. Never mentioned since.

Cray had a contract with AMD, they ditched AMD as quickly as they could in favour of Xeons afterwards.
Quickly dumped.
Cray jumped from AMD for the flexibility and performance of Intel chips | TheINQUIRER

So you want to define the standard compiler flags? There is no end to what you keep making up....

Wiki is quite useful and you are quite useless , you come up with things that you just made up.

AMD sold the chip and you said it did not make sense, so what is it now ? You can not sell things that make no sense, that there are better options is another thing.

You are purposely derailing the thread with so much bullshit.

 

[Shintai]

I agree design costs increase, but that doesn't affect margins, design costs, are not fixed costs, they are variable costs and should never be mixed in with margins.

Having stated that, the total R&D budget of Zen is not just one year, its probably 5 years ago when they started?

So they have enough money to make a decent chip. We aren't talking about beating Intel here, AMD knows why Intel's chips are beating them, they know what changes they have to make, its not something they need to "invent" they just need to figure out how to do it. Over a 5 year span and decent lead they should be able to do that. If they don't, well, AMD is dead. That is it.

Zen is made with what, less than 2/3rds of the Bulldozer budget?

All the next AMD chips that's going to make it better, always ended up selling less than the chip before it. Including revisions.

AMD isn't dead if Zen isn't "attractive". VIA for example is still alive and making profits. And they are in an even worse place.

 

[razor1]

So you want to define the standard compiler flags? There is no end to what you keep making up....

Wiki is quite useful and you are quite useless , you come up with things that you just made up.

AMD sold the chip and you said it did not make sense, so what is it now ? You can not sell things that make no sense, that there are better options is another thing.

You are purposely derailing the thread with so much bullshit.

err why are we talking about compiler flags, that had nothing to do with the blender run. We have no idea of what that blender run consisted of, its better not to even discuss or take it as any proof of anything because we just don't know what they might have done to it.

 

[Shintai]

So you want to define the standard compiler flags? There is no end to what you keep making up....

Wiki is quite useful and you are quite useless , you come up with things that you just made up.

AMD sold the chip and you said it did not make sense, so what is it now ? You can not sell things that make no sense, that there are better options is another thing.

You are purposely derailing the thread with so much bullshit.

The CEO, Lisa Su just told you K12 is dead. Aka ARM for servers is no more with AMD.

AMD sold the chip? Cray was contractual bound to buy it.

 

[razor1]

Zen is made with what, less than 2/3rds of the Bulldozer budget?

All the next AMD chips that's going to make it better, always ended up selling less than the chip before it. Including revisions.

AMD isn't dead if Zen isn't "attractive". VIA for example is still alive and making profits. And they are in an even worse place.

Via didn't have AMD's debt.

 

[Shintai]

Via didn't have AMD's debt.

The debt really isn't that bad, if they cut of the PC part and trim the company. It would relatively quickly be paid.

 

[NKD]

Their current crop of CPU's are on 32 nm, going to 14nm, is going to save them money lol, if you follow the trend of nodes, the cost flattens out as the process matures. 14nm might be more expensive per transistor, but for mm2 over 32nm (comparing same performance chips) its cheaper.

R&D isn't the only defining factor of creating a good product. If you look back into the early days of AMD, they didn't have the R&D budget Intel had, still they were able to do well. Granted they reverse engineered Intel's chips, but still, a smaller budget got them to equal stance with Intel.

Their knowledge of what went wrong with Bulldozer and Phenom, just by fixing those problems they can get a core that can compete somewhat. Its not about reinventing the wheel, they aren't that far behind.......

Also people buying these chips for gaming (low end gaming which I don't think that is the case), won't need an IGP (if this is the case) and right now, these are the only people that are buying AMD CPU's, no businesses buy AMD, at least smart ones, because they know just the power savings alone is enough to go to Intel.

well put. I think they just need to be competitive and its not like they started this business yesterday. You don't need another 100 million in R&D to design a competitive product. We all know where their struggle is in the future and so far its be Global Foundries. We all know they are trying to find a way to lessen that burden. And I agree with their knowledge and where they failed with bulldozer and phenom. I am sure they don[t need R&D budget to see where they went wrong with bulldozer and how it hurt them.

 

[razor1]

Zen is made with what, less than 2/3rds of the Bulldozer budget?

All the next AMD chips that's going to make it better, always ended up selling less than the chip before it. Including revisions.

AMD isn't dead if Zen isn't "attractive". VIA for example is still alive and making profits. And they are in an even worse place.

Maybe but I don't think you can look at it that cut and dry because Zen was started 5 years prior, Bulldozer turn around was fast, with Phenom 2 loosing ground they went at Bulldozer pretty fast, but it was delay after delay that got them. If we take those delays out, which probably were not R&D anymore but trying fix an unfixable product, I think Zen had double the budget.

AMD's goal with bulldozer was to use automated systems to speed up their design process, which failed horribly, so we need to consider, they did speed up their design phase, but the delays had nothing to do with actual R&D.

 

[razor1]

The debt really isn't that bad, if they cut of the PC part and trim the company. It would relatively quickly be paid.

Oh its bad, what it at now 1.7 billion? That is like 1/2 of the company value, this is why no bank will give them money. So if their capital runs dry, game over.

 

[Shintai]

Maybe but I don't think you can look at it that cut and dry because Zen was started 5 years prior, Bulldozer turn around was fast, with Phenom 2 loosing ground they went at Bulldozer pretty fast, but it was delay after delay that got them. If we take those delays out, which probably were not R&D anymore but trying fix an unfixable product, I think Zen had double the budget.

AMD's goal with bulldozer was to use automated systems to speed up their design process, which failed horribly, so we need to consider, they did speed up their design phase, but the delays had nothing to do with actual R&D.

Zen uses the same design methods and is heavily synthetized. Bulldozer was a speed runner design as its main weakness. But Zen tragicomic enough got the same issues as Bulldozer with not being able to clock high enough. Both due to caches and the process. Limiting the Fmax to around 3Ghz. But the downturn started with Phenom. The mantra for both Bulldozer and Zen have become "moar cores.". Shown with the clustered cores and MCM designs.

 

[Shintai]

Oh its bad, what it at now 1.7 billion? That is like 1/2 of the company value, this is why no bank will give them money. So if their capital runs dry, game over.

Because they still bet on PC. Look on what is losing the money.

 

[razor1]

Zen uses the same design methods and is heavily synthetized. Bulldozer was a speed runner design as its main weakness. But the downturn started with Phenom. The mantra for both Bulldozer and Zen have become "moar cores.". Shown with the clustered cores and MCM designs.

I don't think Zen uses the same design methods, I don't think Keller or any sane EE would use automated designs for high performance CPU's, even nV has stepped away form this for their GPU's, and we can see the difference it makes, just doesn't make sense if you want to be competitive.

 

[Shintai]

I don't think Zen uses the same design methods, I don't think Keller or any sane EE would use automated designs for high performance CPU's, even nV has stepped away form this for their GPU's, and we can see the difference it makes, just doesn't make sense if you want to be competitive.

Hand optimizing cost a lot of money. And Keller seems to be more glamour than substance

Nvidia? What did Pascal cost? 2 billion? And that's just a GPU.

You cant do something you cant afford. Also AMD have said nothing in relation to a change in their design methods.

 

[razor1]

Because they still bet on PC. Look on what is losing the money.

That wasn't why they got into to debt, they got into debt for the ATI purchase which at the time AMD was at the top of their game, and the path they were to take was solid, but the CPU division failure along with the r600 was something they were not expecting. Even if they were I don't think they thought it was going to be as bad as it turned out. Who expected the g80? we all thought the g80 was not going to be a unified shader model, I only found out a month or so before launch it was through some nV white papers that were sent to me by a friend.

Who would have thought Phenom (Barcelona) would have its problems on the server end and then the final nail was bulldozer. Its hard to make one mistake like Phenom, but to trump that with Bulldozer damn.

And this is it it was a wrong step to take, but once taken, they have to live with it for 5 years and this is the time when Zen started. With Phenom it wasn't that far behind, i wouldn't be surprised if they are leveraging most of phenom's core IP for Zen, kinda like what Intel did with core tech, they leveraged PIII IP.

 

[razor1]

Hand optimizing cost a lot of money. And Keller seems to be more glamour than substance

Nvidia? What did Pascal cost? 2 billion? And that's just a GPU.

Of course it costs money, but better do it now or never do it again.

I don't know about Keller, but all they need is the right direction. And he should be able to do that.

Pascal they stated 2 billion, yeah but I don't think that is only for the GPU they probably put in everything for their deep learning and other initiatives.

 

[NKD]

Hand optimizing cost a lot of money. And Keller seems to be more glamour than substance

Nvidia? What did Pascal cost? 2 billion? And that's just a GPU.

You cant do something you cant afford.

I highly frickin doubt pascal cost 2 billion. There isn't much difference between pascal and maxwell, nothing major. I am sure it involves other things too. If company is bragging about 2 billion dollar investment I doubt it just gpu alone. Knowing nvidia I am sure it probably means everything based on same design core and all their deep learning stuff. Heck it could be a part of their 2 billion dollar R&D overall for all things they spent over the last few years. That is probably more like nvidia.

 

[Shintai]

Of course it costs money, but better do it now or never do it again.

I don't know about Keller, but all they need is the right direction. And he should be able to do that.

Pascal they stated 2 billion, yeah but I don't think that is only for the GPU they probably put in everything for their deep learning and other initiatives.

But what actually points to Zen being hand optimized and AMD dropping everything they made a standard with both CPUs and GPUs? Besides wishful hopes that is.

Zen is most likely around 4 years now in development btw. Still a good half year till release.

 

[razor1]

I just don't think it would be smart to do the same mistake AMD did with bulldozer, why would they stick their foot in fire twice? If they did, they aren't just stupid, that's like retarded level, mentally challenged, which I can't believe that till I see it.

 

[NKD]

But what actually points to Zen being hand optimized and AMD dropping everything they made a standard with both CPUs and GPUs? Besides wishful hopes that is.

Zen is most likely around 4 years now in development btw. Still a good half year till release.

you think they would spend 4 years on repeating the same mistake they did with bulldozer? I mean they must hate themselves and love wasting money? I highly doubt they repeat the same mistake knowing exactly what was wrong and spent more money not to fix it. lol like Razor1 said they must be border line stupid, and mentally ill.

 

[Shintai]

I just don't think it would be smart to do the same mistake AMD did with bulldozer, why would they stick their foot in fire twice? If they did, they aren't just stupid, that's like retarded level, mentally challenged, which I can't believe that till I see it.

Nonsenses. You cant just get all you want if you cant afford it. You try and do the best with what you have. Synthetized designs wasn't the biggest flaw of Bulldozer, far from it. The biggest problem is they create a product for a segment far out of their reach. The APU for example should be their first product and designed around it. Mobile and desktop with cost as big factor.

 

[Shintai]

you think they would spend 4 years on repeating the same mistake they did with bulldozer? I mean they must hate themselves and love wasting money? I highly doubt they repeat the same mistake knowing exactly what was wrong and spent more money not to fix it. lol like Razor1 said they must be border line stupid, and mentally ill.

The same mistake going for the server segment? Absolutely. But the other parts? No.

SeaMicro is another fine example of a complete "server" loss. That was 400M$ or so down the drain.

 

[razor1]

You have to shoot for the competition's mark or above that, otherwise you won't get anywhere. If you shoot for mid range you end up with Polaris, a mid range chip that's all ya get.

 

[Shintai]

You have to shoot for the competition's mark or above that, otherwise you won't get anywhere. If you shoot for mid range you end up with Polaris, a mid range chip that's all ya get.

What do you think VIA for example does when they start a CPU design? Target Skylake?

Does ARM, Samsung and Qualcomm target A10?

No, they all target something else they can be good at. Be it cost, other features or something else.

AMD got RTG, yet they are pissing it away while they are busy making server CPUs. 99% of Intels consumer sales are CPUs with IGP.

 

[razor1]

They have their niche products, AMD is not trying to go into a niche, where there will be little competition, they are trying to go up against probably one of the few companies that have absolute dominance in their market. But laying the foundation for future, better products is just as important too. Because to sustain a business model, just in case they can't hit the top end, they still need to have competitive products that aren't going to burn a hole in their pocket when they sell them.

Qualcomm is trying to target high performance CPU's as of late, I don't know how well they will do but we will see.

 

[NKD]

What do you think VIA for example does when they start a CPU design? Target Skylake?

Does ARM, Samsung and Qualcomm target A10?

No, they all target something else they can be good at. Be it cost, other features or something else.

AMD got RTG, yet they are pissing it away while they are busy making server CPUs. 99% of Intels consumer sales are CPUs with IGP.

That's not where amd competes. They are not in mobile market they can't just drop everything and go in to a highly competitive market and crowded one on top. They have been primarily cpu company and they have one thing to do make a competitive product that isn't a failure like bulldozer was. All they need to do is lay a foundation. let's say zen is big improvement over excavator, they can build off that, they can change fabs down the road which they might be planing and go for higher clocks on zen+ if global foundries is holding them back there. Lots of options but first step is building something decent and build off it. That's all. They don't need Intel killer they need a decent product that competes well and way above their shitty current offerings.

 

[razor1]

I don't think having their own fabs is a good idea lol, fabs are expensive, just retooling a fab for a new process is now in the few billions of dollars. But the rest of it yep sounds good.

 

[NKD]

I don't think having their own fabs is a good idea lol, fabs are expensive, just retooling a fab for a new process is now in the few billions of dollars. But the rest of it yep sounds good.

Absolutely thats what I meant. Just going to tsmc if they have to. By no means I mean new fabs lol, never. They can't afford that shit, lol.